Evaluate Metal Segregation in Gold Bullion

Evaluate Metal Segregation in Gold Bullion

Several years ago the writer was connected with the Mint and Assay Service of the Federal Government as Assistant Assayer at the Salt Lake Assay Office. At that time cyanide bars formed approximately half of the bullion purchased by that office. Disagreement in valuation between the producer and the office was not infrequent and to a lesser extent between the Assay Office and the Mint, this latter issue being soon obviated, however, by adopting uniform methods of sampling the bars. A nice problem seemed to offer itself for research and the writer began a series of experiments which were soon terminated, however, by his going into other work, and which he regrets have not since been completed. The results attained may suggest lines of approach to this problem and are, therefore, offered in this article.

In the discussion of these papers, comparison has been made in terms of copper bullion, but to the writer this seems of questionable value. Copper bullion may resemble silver bullion, but its similarity to gold bullion with little or no silver content is another thing. The freezing-point diagram of Roozeboom, is for the gold-silver series as stated. The original problem under discussion, however, is not this alloy but gold with base metals other than copper. As for the statement that the corners are the worst places or positions to drill: In the writer’s opinion there is not much choice, if the outer film of approximately 1 in. be excluded. A glance at the diagrams accompanying this article will show that a wide variation from the dip average is apt to be encountered at any of the places drilled in these experiments.gold bullion

Sampling Methods

A few details of manipulation will be reviewed here to throw light on the discussion. The large bars are cast in molds 12¼ by 5¼ in,, and range from 2 to 2½ in. in thickness. They are essentially prisms, the mold being tapered slightly to permit easy removal of the bar.

Three methods of sampling are in vogue:

  1. Chipping diagonally opposite corners of the bar;
  2. drilling the bar;
  3. dip-sampling as the bar is being poured.

Methods (2) and (3) are the ones commonly employed in sampling cyanide bullion. Drilling is done as shown in Fig. 1.

The two samples of top drillings are mixed and assayed as the top sample, and the two bottom drillings are mixed and assayed as the bottom sample. If the bars are not uniform in composition and if the variation is gradual it is readily seen that the top sample really represents, an average of the top and center, or approximately a plane halfway between the top and the center of the bar. Similarly the mixture of bottom drillings represents a plane halfway between the center and the bottom. Now an average of these two planes is not necessarily an average of the whole bar, and where segregation is the rule and not the exception, this method may be open to criticism. This point, I believe, has been somewhat overlooked or neglected, and may well be worthy of experimental investigation. As emphasized by Mr. Dewey, these drillings themselves may be rather heterogeneous, the smaller fragments differing appreciably from the coarser in gold content, and this may result in wide discrepancies in assays of the same drill sample.

Dip samples may be taken as follows: One sample may be taken from the top of the melt just before it is poured into the mold, or a portion of the stream cut out just after pouring begins. This represents the top of the melt, or the bottom of the bar. A second sample taken from the pour just before completion represents the bottom of the melt, or the top of the bar. These samples are the most reliable, if the melt was thoroughly stirred just before pouring, and should check very closely within themselves and with each other.

Variations in Assays

Irregularities in the assay of cyanide bullion seem to have been first noted by Edward Mathey, who thus summarizes some interesting and instructive experiments:

  • (a) Alloys of gold with base metals, notably with lead and zinc, now largely met with in industry, have the gold concentrated toward the center and lower portions, which renders it impossible to ascertain their true value with even an approximation to accuracy;
  • (b) when silver is also present, these irregularities are greatly modified. The method of obtaining “cooling curves” of the alloys shows that the freezing points are very different where silver is present and where it is absent from the alloy;
  • (c) this fact naturally leads to the belief that if the base metal present does not exceed 30 per cent., silver will dissolve it and form a uniform alloy with gold;
  • (d) this conclusion is sustained by experiments (described in his article) which, in fact, gradually lead up to it and enable a question of much interest to be solved.”

In another series of experiments with alloys of gold and platinum, Mathey clearly demonstrates the tendency of the latter metal to liquate toward the center of the mass while cooling, Dewey’s articles call attention to the existence of variations in assays of such bullion and the possible causes, but he does not show the nature of the variations.

Mention is made in metallurgical texts of the solvent property of silver where certain base metals, such as zinc, lead, etc., are present. The writer has had occasion to confirm this point in connection with the assay of gold bullion. If there is considerable silver present, 60 points (6 per cent.) or more, a uniform bar is generally obtained, but where the silver content is slight, segregation or liquidation may take place on cooling and the resulting bar may be far from uniform in composition. With these points in mind the writer assayed a number of drillings taken from various portions of bars. Grouping these results in various ways a number of interesting and suggestive diagrams may be constructed which illustrate graphically the lack of homogeneity in some cyanide bullion.

Qualitative tests on some of the bullion indicated the presence of small amounts of mercury, lead, bismuth, arsenic, antimony, zinc and chromium. Dewey reports cadmium, nickel, iron and copper from some similar bullion. The silver content ranged from zero plus up to 10 points (or 1 per cent.) in rare instances. As a rule these bars are very brittle, and upon fracture show some crystal faces well developed, and a variety of colors. These ranged from the color of marcasite and pyrite to those of chalcopyrite. The faces of the bars were characteristically marked with small gas blebs, giving them a pitted appearance. This was especially noticeable along the upper four edges.

Now when diagonally opposite corners of such a bar are chipped and assayed, the values obtained may have a wide variation among themselves, and in most cases are several points below the values shown by dip samples. Hence agreement between chip samples and dips from such bullion is hopeless, and the greater the number of duplicate assays made, the more firmly established is this difference. When this feature was recognized the mint issued the following instructions for sampling such bullion: “A drilling is to be made about one inch from each edge on two opposite corners of the top of the bar, and should extend half way through the bar. These two drillings are to be thoroughly mixed and assayed as the top sample. The bottom sample is to be taken in the same manner, corners being chosen not under the top corners drilled. (This is illustrated in Fig. 1.) Dip samples are to be taken as previous^ described and assayed with the drill samples.”dip-samples

In Table I are given some results obtained from assaying such samples. Where only one assay value is recorded, as for the top dips of bar 18, it should be understood that the other assay was spoiled, and not that it was too far off to put on record. Similarly for one of the top drills of bar 41. Eighteen of these bars were reported as having no silver (meaning less than 0.05 per cent.), the maximum silver content being 12 points (1.2 per cent.), whereas the average for the 48 bars was less than 3.8 points (0.38 per cent.).

Comparison of the assays of the top and bottom drills shows that in 38 cases the top drillings assay higher than the bottom drillings, whereas in 10 cases the reverse is true. In many instances a considerable range of assay values appears. Examining the dip assays in the same way it is found that in 33 cases the top dips assay higher than the bottom dips, and in 12. cases the reverse is true. In three bars the top and bottom dips average the same. As a rule, however, there is less difference between the top and bottom dips than there is between the top and bottom drillings.



Comparing the assays of drillings and dip samples it is seen that in 28 instances the drillings assay higher than the dips; in two cases they are alike, and in 18 cases the dips assay higher than the drillings. In nearly every instance the drillings show greater variation than do the dip samples, and yet in some samples (Nos. 1, 2, 12, 20, 28, 30, 31, 41, 43, 47, and 48) there is a marked variation here, even in dips from the same portion of the melt.

The molds are of cast iron and are heated to about 150° to 200°C. before the bullion is poured into them. As the metal is a better heat conductor than the air, that portion of the bullion which is in contact with the mold probably solidifies first, whereas the top cools last. This might produce a bar of the following characteristics: Since the four lower corners chill first, they might contain a higher percentage of the metals and non-metals of the highest freezing points. By the term “freezing point ” as here used is meant a resultant of the absolute freezing point of the metal or alloy, and the saturation point of this same metal or alloy in the melt. This is offset to a certain degree by some of the metal solidifying as soon as it strikes the mold and not permitting any segregation. Next in order would be the four upper corners, then the bottom and sides, and lastly, the top which may be somewhat protected or blanketed by a thin coating of slag. The middle of each face would probably freeze later than any other point of that face, and the center of the bar, approximately, would solidify last, or possibly a thin film through the center and parallel to the bottom might be the last to take the solid form. The specific heats of the various elements making up the bar would exert a differential tendency in its cooling. If mercury were present, on account of its low melting point, it might tend to segregate to that portion of the bar which freezes last. The solvent properties of the metals and their alloying characteristics would also exert an appreciable influence here. Arsenic is known to cause a lack of uniformity, even where present in small amounts, but the exact nature of this action does not seem to be understood. A considerable amount of the volatile and easily oxidized elements are driven off during the melting, stage, zinc, lead and bismuth being detected in the flues, but as charcoal covers are used, this expulsion is not complete. Unfortunately, an elaborate qualitative and quantitative analysis of this bullion was not feasible, owing to lack of equipment, and to certain restrictions placed upon the operating departments of the assay offices. Extensive drilling was therefore resorted to for a few bars, and some interesting variations determined.

From the drillings and dips taken in the usual manner for bar 49, the following assay values were obtained:


This bar was specially drilled as indicated in Fig. 2, each drill hole extending to the middle of the bar. The bar was 10½ in. long, 4½ in. wide and 2½- in. thick, and weighed 824.81 troy ounces.


The points x, x represent the regular places for drilling the top of the bar, and y, y the places for taking bottom drillings. The points a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p are on top of the bar and spaced as shown. The points q, r, s, t, u, are on the bottom of the bar. The drillings from


each hole were carefully mixed to insure uniformity (as far as possible) and were assayed in duplicate. Bars 50 and 51 were drilled in the same manner, and all the samples likewise assayed in duplicate. The values determined from these assays are given in the accompanying tables.


From the drillings and dips taken in the usual manner for Bar 50, the assay values shown in Table V were obtained.




From the drillings and dips taken in the usual manner for Bar 51, the assay values shown in Table VII were obtained.



A large number of drillings were taken from two other bars. The method of drilling is shown in Fig. 3. The dimensions of the bars were: length, 12¼ in., width, 5¼ in., and thickness, 2¼ in. Bar 52 weighed 949.34 oz., and Bar 53 weighed 942.34 oz.

Drillings from a to b’ through z, inclusive, were taken from the top of the bar. The bottom drillings c’ and d’ were taken under y and d, respectively. After assaying samples a and b’ separately, the samples were carefully mixed and assayed as the top drills. Similarly, c’ and d’ were first assayed separately and then mixed and assayed as the bottom drills. Separate treatment of the end drillings shows different values for the two ends of the bar, a point lost .sight of where the drillings are first mixed and then assayed. This becomes more noticeable when diagrams are constructed on these values.


From the top drillings obtained by mixing samples a and b’, the bottom drillings obtained by mixing samples c’ and d’, and the dips taken in the usual manner for Bar 52, the assay values given in Table IX were obtained.



Drillings c to w inclusive (Fig. 4), were taken from the bottom of the bar. Samples a and b were drilled from the top of the bar over e and u respectively. After assaying samples a and b separately, the samples were carefully mixed and assayed as the top drillings. Similarly, c and w were first assayed separately and then mixed and assayed as the bottom drillings. As was true with Bar 52, comparison of assays a, b, c, e, u and w shows considerable variation between the two ends of the bar.


From the top drillings obtained by making samples a and b, the bottom drillings obtained by mixing samples c and w, and the dips taken in the usual manner for Bar 53, the assay values given in Table XI were obtained.


Construction of Diagrams

The horizontal length of the diagram in each case represents the outside dimension of the bar along the line of the profile. As no sample is taken less than an inch from the edge of the bar, the curves all terminate inside of the diagrams. The vertical control is that of fineness, the extremes taken so as to include all assay values on the bar. This feature is apparent on the last or composite diagram for each bar, upon which are assembled the assay values of all the drillings from that bar. In this diagram the values are all projected upon a plane passing lengthwise through the center of the bar and normal to the end and bottom faces.

The light solid horizontal line passing through the diagram—and curves in some cases—represents the average value of the dip assays, and probably is a very close approach to the true value of the bar. The


letters along the base of the diagram are the points on the bar from which the drillings were taken, and over these letters are placed the corresponding assay values. Each diagram is thus an assay profile along a line through the lettered points, the last diagram for each bar being a composite profile along a longitudinal medial plane, all assay values of the drillings being projected upon this plane. All drilling assay values are placed upon the diagrams except Plates XII and XIII, and the curves are drawn through the average for any point.

Several methods of drawing curves through these average values are possible, and each one may be nearly equally justified.

  1. Straight fines may be drawn between consecutive average values, the profile consisting of a zigzag fine made up of straight segments.
  2. A smooth curve may be drawn through consecutive average values, without regard to what might lie beyond each end of the curve.
  3. A smooth curve may be drawn through consecutive average values, and this curve be bent toward the horizontal at each end.
  4. A smooth curve may be drawn through consecutive average values, but warped toward the average dip value. This would assume all other than dip values as departures from the true value and as abnormalities.
  5. An average of two consecutive averages may be taken and its departure from the dip average noted (call this departure d). Now take the dip average as a datum line, and half way between the two averages considered (horizontally), insert d with its sign changed. This is merely a device for obtaining low values to offset the high values. Either short straight lines or a smooth curve may then be drawn through all points, average values and the new points thus computed.

The first method assumes that assay values were taken at all maximum and minimum points, and that the gradation between points is uniform. This is certainly not necessarily true, and is probably never attained.

The second method considers that all maximum and minimum points are represented, but does not suggest uniform gradation between points sampled. Thus the curve between points is dependent upon adjacent


values as well, and therefore is probably in less error than the zigzag line of method (1). As it is not necessarily true that all maxima and minima are represented, there is undoubtedly a departure from actual conditions, but the only remedy for this would be a determination of essentially all points of the bar, a proceeding manifestly impossible.

The third method bends the curve toward the horizontal at each end. This undoubtedly possesses merit in some cases, but may introduce errors in other instances. It is essentially the same as (2).

The fourth method brings out interesting contrasts, but certainly cannot represent true conditions in the bar. It might still leave all or most of the profile above or below the dip average, a lack of balance probably not the case. A glance at the assay values and diagrams for bars 49, 50, and 51 will show the difficulty of realizing such conditions with some bars.

Such a diagram as outlined in the fifth case or method would show some of the variations demanded if we are to consider the dip average as the true value. Although introducing mid-points, in the opinion of the writer it may more nearly meet the conditions of the case than any of the other

bar 50

methods. In all probability it does not properly place these variations, but at the same time it shows what the minimum variations must be if the dip value is the true datum line. For instance, such a composite diagram as Plate VI, B, cannot tell all the story, for where are the low values to offset the high ones and give the true bar value? Much of this adjustment is undoubtedly accomplished in the outer film of the bar, outside the curve terminals, but not necessarily all.

In the first nine plates the second method was used. For bar 53 the third method was followed, and it shows curves nearly identical with those obtained with the second method. In Plate XIII the first and fifth methods were used.

Discussion of Bullion Assay Curve Data

Bars 49, 50 and 51.—As these three bars were of the same size and shape, and as they were sampled in identically the same manner, a comparison of their diagrams may be instructive. Noting the curves along the diagonal line on top of the bar, it is seen

bar 51

that essentially all the average values are above the average dip value, emphasizing the difficulty of obtaining the fineness of such bullion from drillings. Another feature is the large bulge or wave-like crest in the curve near one end. The reason for this is not known, as the melt is poured across the mold from end to end, and principally at the center. As the drillings from each end are mixed and assayed in duplicate, both end values are alike. Duplicate values would not likely result if the drillings from each end were assayed separately. In the case of each of these three bars the curve is convex upward.

Examination of the curves for assay values along the central line on top of the bar shows a similar wide range of values and a still greater departure from the average dip value. Each of these curves is essentially concave upward, in contrast to the convexity of the curves along the diagonal line. In speaking of the cooling of the melt, mention was made of the probability that the center of the top of the bar may be the last part of this face to solidify. If this is true in principle for the bar itself,

bar 51-2

and a metal such as mercury were present, a lower assay value for this portion of the bar might be expected. However, it might be expected to fall below the average dip value rather than above. As in the previous case, nearly all the assay values lie above the average dip value.

The curves for the central line along the bottom of the bar have one noticeable trait in common; namely, a concavity upward which is due to a lower value near the center of the bar. In keeping with this feature is the hypothecated action due to the unequal rates in cooling. The center of any face probably cools or solidifies later than any other portion near the edges, such as along the borders or near the corners. At the centers, these curves, except that on Plate VII, A, pass below the average dip value, although most of the curve is above.

bar 52

In the composite diagrams the curves are still more irregular but all illustrate the high values of drillings and the inadvisability of attempting to determine fineness of such bullion from such samples.

In Plates II, B, and IV, B, the end averages are nearly coincident with the dip averages. In Plates VI, B, IX, D, and XII, B, however, these end average values are not the same as the dip averages, but are higher in some cases and lower in others.

Bars 52 and 53 were sampled according to a different scheme. Lines were drawn parallel to the edges, and drillings were taken at short intervals along these lines (see Figs. 3 and 4).

bar 53

The bottom of bar 52 and the top of bar 53 were sampled in the official manner previously described. Plotting these values lengthwise and crosswise of the bar, two sets of curves were obtained. A composite diagram is also given for each bar. The diagrams are arranged in pairs so as to contrast similar portions. Thus for bar 52 the profile along aeimquy corresponds in position to that along dhlptxb’. Each curve connects a series of averages for assays of drillings taken parallel to, and at a given distance from the edge of the bar. On a given face these profiles

plate ix bar 52

might be expected to show similar variations if the liquation in such cooling is merely a matter of rate of cooling rather than a complex function depending upon many variables, conspicuous among which is apparently composition. Possibly the term liquation is here inapplicable, as the solid bar may (and in all likelihood does) somewhat resemble a rock of heterogeneous mineral composition, rather than a liquated magma. Close proximity to the walls of the mold may exercise much the same function as proximity to inclosing wall rock in the case of a cooling magma. Here, however, the solution of metal in metal (or non-metal)

plate x bar 53

and the peculiar properties of alloys may vary the final products considerably.

Thus in Plate VII, A and B, each curve shows a conspicuous crest at the center and a conspicuous trough, to the right of this crest. Or the right half of the curve shows two crests with an intervening trough. In each case the central crest separates two dissimilar curves. Neither curve is decidedly concave or convex upward. Unlike preceding curves they are approximately bisected by the dip average.

Comparing C and D a slight similarity is noted, in that both are convex upward and the variation from one edge of the bar to the opposite

plate xi bar 53

side is consistent for both profiles. Considerable variation, however, is shown in these profiles.

In Plate VIII, contrasting A and B, some like features are apparent. The curves are both concave upward and average well above the dips. The central trough in each reaches nearly the same minimum value, whereas the ends show similar maxima.

C and D are both convex upward and average well above the dips. These curves which are convex upward, are just the opposite of A and B which are concave upward, and show that if there is such a thing as concentration of a metal such as mercury toward the center, it is more than offset by some reverse process, at least locally.

In Plate IX, A and B are not much alike. A is unlike C on Plate VIII, although it is an adjoining profile, whereas B is similar to D of Plate VIII, but less pronounced. C, Plate IX, is concave upward and

plate xii bar 53

in harmony with a possible concentration of such a metal as mercury at the center. At variance with this, however, is the fact that the entire curve is above the dip average.

The composite diagram D shows a curve with three well-defined crests and two troughs, and with an average somewhat above the dip average. Although the range in individual values is considerable, the point averages are fairly close to the dip average. The grouping of the assay values shows well-defined maxima and minima points.

Bar 53 was drilled in a way somewhat similar to bar 52 except that the drillings were from the bottom of the bar instead of the top. This, it was hoped, might furnish data for contrasting cooling in a face exposed to the air or to a thin slag coating, as compared with the cooling of a face next to the mold. Drilling one or several of the lateral faces might show other gradations and bring out interesting contrasts.

In Plate X, A and C are similarly situated, and both show downward

bars 52 and 53

gradients from left to right. Neither one, however, is decidedly concave or convex upward, and the curves are not very similar. Curve A shows erratic tendencies whereas C indicates a less variable decline in value from left to right. B, which is along the medial line of the bottom of the bar, is concave upward, but is in decided contrast with both A and B. In some ways, however, it seems to be a composite of both A and C, except for the values of s and v which seem abnormally high for this bar.

In Plate XI most of the curves are convex upward, but do not go together well in pairs. For instance, A and B are similarly situated, but are in strong contrast; similarly for C and D, C being slightly concave upward. Since E is on the same side of the bar as A and C we might compare them. It is not very far from being a composite of A and C, although the left end is rather low.

Comparing F with B and D we find that the convexity upward of B, which is more conspicuous in D, has nearly disappeared in F, the maximum point having fallen well below the dip average.

A of Plate XII is a fair composite of E and F on Plate XI, and lies between them on the bar. It is nearly a straight line, the variation being a slight concavity upward. B on Plate XII is the composite for bar 53, and, although it shows wide range between individual assays, indicates a bar more nearly uniform than any of the preceding. The curve is slightly concave upward although the average value of drillings is very close to that of the dips.

In the diagrams on Plates X, XI and XII, the curves are bent toward the horizontal at the ends, but this feature is inconspicuous in the final product.

If a curve were constructed according to the fourth method suggested at the beginning (warping all portions of the curve between the maxima and minima toward the dip average) the result would be rather striking and extremely improbable. This would be especially conspicuous for bars 49, 50 and 51.

For these reasons two diagrams are given for composite values of bars 52 and 53, constructed according to the first and fifth methods, the zigzag line of short straight segments being used rather than the smooth curve. The composite diagrams for both of these bars show average points which, are less extreme, measured from the dip average, than those for bars 49, 50 and 51, and for these reasons were chosen. The intermediate and extreme points are chosen so that the areas above the dip average but below the profile are equal to the areas below the dip average but above the profile. Or considering the areas between the dip average and the profile as positive or negative according as they are above or below the former, the positive areas equal the negative. Some such balance must exist if the dip average represents the true assay value of the bar. Such a diagram does not show, necessarily, the exact location or manner of such variations, but it does give a minimum variation. Possibly much of this balance may be accomplished in the outer inch of the bar, rather than within this profile. Complete adjustment in this outer portion, however, seems improbable to the writer, especially in profiles such as Plates II, By IV, B, and VI, B. In profiles such as Plates IX, D, and XII, B, where values in the profile are both below and above the dip average, it seems much more reasonable to suppose that an approximate balance is maintained within the profile. The same may be true for Plates II, B, IV, B, and VI, B, in which cases the local variations are extreme in places. Although the gradation is almost surely not along straight lines, the turning points may be as sharp as those in the diagram, and are not necessarily curved crests or troughs.

In both profiles on Plate XIII one end of the diagram shows a rising value, whereas the other end shows a decreasing value. Neither extreme is far from the dip average but the erratic variation is thus shown. These two bars, however, are less heterogeneous, or variable through a smaller range, than many cyanide bars, and although an average of all the drillings is close to that of the dips in these two cases, it is decidedly otherwise in many other instances, such as bars 49, 50 and 51, as well as many of the bars listed between 1 and 48.

Another interesting feature brought out by these diagrams is the variation in assay value of individual samples. In the simple diagrams this variation seems equally conspicuous along the profile (or bar) and does not appear to favor any single portion of the bar. The composite diagrams cannot of course be here considered, as there is superposition of points, due to projection on the medial plane.

In a study of these curves a few characteristics seem to be common, but no attempt will be made to offer other than tentative reasons for such. It is to be hoped, however, that a further detailed examination may be undertaken to clear up these points. Such an examination is of necessity confined to the government mints (or possibly assay offices) but for some reason the men in these places have been handicapped by a lack, either of time or of a desire for such an investigation. Considerable extra time is required for such detailed sampling, and still more elaborate methods must be pursued before a satisfactory solution is reached. This is out of the question in the assay offices as the time element is an important feature, and also on account of the bookkeeping system used, no sampling being possible after settlement on a bar is made. At the mints an unusual opportunity is afforded, and such work would be of considerable value to the service as well as of scientific interest.


  1. Drillings taken near an edge are extremely variable, but generally assay higher than the dips, a fact which may be due to early solidification and an accompanying segregation in those portions of the bar of metals of the highest freezing points, taking into account mutual solubilities and alloying tendencies. In a platinum-gold alloy this same characteristic is very apparent, but is due, as Mathey observes, to the liquation of the platinum.
  2. The center of the bar, top and bottom, is usually lower in value than the dips. As suggested early in this paper, this may be due to a late solidification in these and similar points, which results in a higher percentage of base metals with low freezing points. Mutual solubilities may be a factor here, also, but the substance in large excess (gold) does not concentrate toward the center or last place of solidification as does quartz in the cooling of a quartz-rich magma.
  3. In a majority of cases the top drillings assay higher than do those of the bottom, and the same is true of the dips. In order to balance the personal equation the method of weighing up assays was reversed so that the weigher who received the top samples at first, later received the bottom samples. For this reason samples were frequently exchanged in weighing. The same variations, however, continued to be apparent.
    An extreme range in the dip assays is in some cases due to the presence of impurities introduced while the dips are being taken. Where a scorifier is used to catch part of the stream of metal, small pieces of clay might easily be caught and retained in the bullion granules.
  4. Drillings taken at anyone point along these profiles vary as widely in assay value as drillings of any other point similarly, chosen. Also, maxima and minima points are as apt to occur near the center of the bar as near the ends. From this standpoint the corner drillings seem no better nor worse than drillings from nearer the center.
  5. The liquation in some cyanide bars is similar to that which takes place when the rare metals are present in appreciable amounts, but in the former instance it seems to be due to the presence of certain base metals.

Attention is again called to the fact that these difficulties are not met with where a small amount of silver is present, as the presence of this latter metal seems to diminish segregation. Lessened segregation is probably due to the solvent properties of silver toward certain base metals such as lead, zinc, etc. Silver may exert an appreciable influence on other metals or non-metals also. Where 90 or 100 points of silver were known to be present in the bar, no dip samples were taken, and the drillings rarely failed to check well within the limits allowed. This is doubtless due to the fact that silver or silver-gold alloy in which the silver content is nearly 10 per cent., or better, is a good solvent for the metals or non-metals causing the segregation. Pure or silver-free gold is not a good solvent for these substances and hence does not prevent the formation of an irregular bar.

This suggests one way to overcome the difficulty of assaying cyanide bullion, although it does not solve the problem. If to all such cyanide bars known to contain little or no silver, a small amount of the metal be added in melting, a uniform bar after melt could be easily obtained and its fineness accurately determined. From the fineness thus obtained the inquartation silver could be subtracted, leaving the silver fineness of the original bar. As such gold bars are alloyed with silver for refining operations, the above suggested operation would mean a little additional work for the computers only, but this extra work would be more than offset by the increased accuracy of the gold assay.


Frederic P. Dewey, Washington, D. C. (communication to the Secretary)—I wish to congratulate Professor Hance and to thank him for his extremely valuable addition to the fund of data upon this subject. It presents clearly conclusive and incontrovertible proof of the unreliability of drill samples of this class of metal for determining the value of such bullion. I am afraid, however, that his hope that someone in the mint service might continue such an investigation upon a more elaborate and exhaustive scale is, for the present at least, doomed to disappointment.

As a practical matter, the mint service is not now particularly interested in segregation in crude bullion. So much trouble and friction arose, both in the original purchase and in the transfer from one institution to another, in handling segregated bullion, that it became necessary to adopt drastic measures and now it is seldom attempted to determine the value of such bullion in the service. When bullion known or even suspected of being segregated is presented for purchase, it is at once strongly refined in the pot and the disturbing elements so far removed that concordant assays may certainly be obtained from the dip samples and a reasonably fair agreement obtained from drill samples. The mint service prefers that the owner should refine his bullion himself, but if he does not we will do it at his expense and return the slag to him when desired.

Much work has been done in the mint service trying to arrive at the value of deposits of segregated bullion without strong refining. It was no unusual thing for such deposits to be melted three times at the purchasing office and to have 30 to 40 assays on the metal made before buying it. Yet when shipped to a mint it was found necessary to melt two or three times more and to make 15 to 25 assays before the mint would accept the bar, and there were considerable differences between the valuations of the two offices. In what was undoubtedly the worst case ever investigated, the purchasing office melted the deposit, 953 oz., three times, with a loss of 105.5 oz., and made 46 assays; the mint melted three times, with a loss of 6.6 oz., and made 28 assays. The mint allowed the purchasing office 1.178 oz. of fine gold more than the purchasing office claimed, while an extensive investigation of the mint samples by the Bureau, in which over 100 independent assays were made in various service laboratories, indicated that the metal carried 0.7 to 0.9 oz. more of fine gold than the mint allowed. Preliminary assays on the original metal indicated that it was about 661 fine in gold. Even after all the assays made on the mint samples, I would not be absolutely positive as to the composition of the final bar, but it may be taken as close to 751 fine in gold, 60 in silver and 189 in base. Undoubtedly more refining was required in order to obtain thoroughly reliable and satisfactory assays. Subsequently a bar from the same mill was refined to over 900 fine in gold by the purchasing office and accepted at the mint, on six assays, at the purchasing office valuation.

One office expended a tremendous amount of energy upon this question. For six months it was attempted to determine the gold in all such deposits, several hundred of them, as received, for the purpose of comparison with the amount paid for after melting and more or less treatment by the office. There is a sprinkling of cases where the difference was within reason, $10 either way, but a very large proportion of them showed a much wider variation. Some of them were startling and these were not by any means all on one side. In nearly 100 cases, selected at random and tabulated for an entirely different purpose, there were five cases of plus differences and six cases of minus differences of over $200 each. The highest plus variation was $609.43. The highest minus variation was $413.76. Owing to the many and serious difficulties of determining the value of the metal as received, it would not be at all proper to speak of these differences as gains and losses. The best that can be done in this line is to call them apparent gains and losses, but even this is not always entirely warranted.

In some cases the metal was received in small bars and several melted together, but this by no means secured agreement in the two valuations. Six bars from 79.78 to 157.01 oz. in weight were melted together, with a loss of 10.63 oz., but the assays varied from 349.9 to 352.6 fine in gold, and the bar was remelted, with a further loss of 43.13 oz. when the gold assays showed from 371.3 to 371.9 fine, upon which the bar was reported at 371.5 fine in gold, the silver being 557 fine and the base 71.5. Four assays had been made on each one of the small bars, making a total of 24, but there was an apparent loss of $201.51 shown by the final bar. The maximum plus difference noted above was on a mass melt of three bars, about 200 oz. each, totalling 609.83 oz., which was melted with a loss of 73.57 oz. The refined bar showed 908 fine in gold, 62 in silver and 30 base.

Twelve small bars weighing 39.70 to 50.15 oz. each, and totalling 548.38 oz. were melted together and held under blast for 4.5 hr. with a loss of 21.93 oz. in weight and an apparent loss of $67.54, but the gold assays varied from 917.9 to 919.9 fine. The bar was remelted, with a further loss of 7.17 oz., but with an apparent gain of $8.87 in value over the first melting. There was an apparent loss of $58.67 on the calculated value of the 12 individual bars. The final bar showed 932 fine in gold, 48 silver, and 20 base.

Four bars ranging from 156.88 to 223.07 oz. and totalling 751.10 oz. were each assayed four times showing wide variations in the sets of gold assays. Roughly speaking, the silver could be said to average approximately 48 fine and the base 130. Upon being melted together, with a loss of 37.29 oz. the gold assays ranged from 863.2 to 865.7 fine. The apparent loss in value was $7.69. On a second melting with a loss of 22.66 oz. the gold assays varied from 892.4 to 894.4, and there was an apparent gain of $2.42 on the figured value of the four individual bars. On a third melting with a further loss of 21.21 oz. the gold assays ranged from 917.6 to 918 fine, and there was a final apparent loss of $47.92 on the estimated value of the four bars. The final bar was 55 fine in silver and 27.5 base.

After this investigation the office abandoned all attempts to nurse along such metal by mild treatment in stages and proceeded at once to strongly refine all suspicious bars. In general, the method consists in blowing air onto the surface of the molten metal, adding an acid flux as necessary, thickening occasionally with bone ash and skimming.

Recently the shippers have supplied their assays of 14 bars for comparison. Curiously the assay office results showed gains of gold over the shipper’s figures in seven cases and losses in seven cases. As received, these bars varied in weight from 800 to 1,100 oz. with a total of 13,122.72 oz. After refining they weighed 736 to 1,043 oz. totalling 12,125.83. The total loss in melting was 996.89 oz. or an average of 71.2 oz. per bar. The individual loss in melting ranged from 50.75 to 89.40 oz. per bar. As received they varied from 240.5 to 301.1 fine in gold, 556.3 to 628.9 in silver and 122.5 to 173.5 in base. After refining, the apparent gold gains ranged from 1.519 to 5.601 oz. on a bar, the total apparent gain being 22.128 oz. The apparent gold losses varied from 0.932 to 4.485 oz. on a bar and totalled 15.659 oz. The net apparent gain of gold was 6.467 oz. In the case of the silver there were 12 apparent gains ranging from 1.44 to 11.45 oz. on a melt and totalling 64.58 oz. The two apparent silver losses were 3.33 and 23.78 oz. The net apparent gain of silver was 37.47 oz. The slags were returned to the depositor and undoubtedly he realized further gains from them. These figures appear to show a substantial gain to the shipper by having his bullion strongly refined before being purchased by our assay office, but they emphasize most emphatically the difficulty of correctly determining the amount of gold present in the original unrefined bullion.

A great many of our troublesome bars have been over 100 fine in silver and some of them very much over. Eight bars ranging from 43.42 to 138.04 oz. and totalling 702.30 oz. were each assayed for gold four times. Naturally, the range of the assays on the smallest bar was not so very much, being only from 275.4 to 276.1, but on the largest bar it varied from 282 to 288.4. The silver estimates ranged from 570 to 620 and the base from 84.5 to 144. On melting these bars, with a loss of 52.89 oz., the gold assays ranged from 306.3 to 309.3. It is scarcely worth while to compare the value of this bar with the computed values of the individual bars. On remelting the bar, with a loss of 28.5 oz. the gold assays ranged from 322.1 to 323.1, with an apparent gain of $8.19 in value. On a second melting, with a further loss of 3.5 oz. the gold assays ranged from 324.3 to 325, with a further apparent gain of $2.26. The final bar was 661 fine in silver and 14.5 in base.

Nine other bars from the same mill totalled 670.63 oz. The mass melt weighed 644.51 and the gold assays ranged from 277.6 to 282.3 fine. On remelting, with a loss of 52.22 oz. the gold assays ranged from 300 to 301.4 fine. On a second melting, with a further loss of 15.93 oz., the gold assays ranged from 308.4 to 309 fine. The final bar was 672 fine in silver and 19.5 in base.

There is, of course, no comparison between this metal and that so successfully treated with silver by Professor Hance, but seven small bars totalling 270.47 oz. yielded a mass melt weighing 250.40 oz. on which the gold assays varied from 742.5 to 745.6 fine. On remelting, with a further loss of 14 oz. the gold assays varied from 786.8 to 789.5 fine. On a second remelt, with a further loss of 17.35 oz. the gold assays varied from 846.9 to 847.8 fine. The final bar showed 100 fine in silver and 53.5 base. A bar weighing 726.93 yielded four gold assays varying from 807.4 to 811.1, the silver being approximately 115 and the base 76. On melting, with a loss of 42.37 oz. the four gold assays varied from 855.7 to 855.9, the silver fineness was 120 and the base 24.5.

It is therefore apparent that much more investigation is required in order to determine the class or classes of gold bullion in which segregation can be overcome or reduced by the addition of silver.

George C. Stone, New York, N. Y.—I do not know whether the gold people have tried the method used by the zinc men in sampling. We have found that drilling was very unsatisfactory. Ordinary spelter segregates pretty badly; the lead is irregularly distributed, and there is always more at the bottom of the slab, so that when we drill we drill completely through, but we have found that the much more satisfactory way with spelter was to saw the slab, and by taking saw cuts from each side more than half way across, we obtained samples that would check with each other, and give a most excellent material for the laboratory. The sawdust is so fine that it is easily mixed, and several duplicate samples can be weighed out and checked with each other perfectly, and samples taken by different people will check extremely well.

I have never heard of it being used in connection with gold bullion, but I think it would be worth experimenting with. Some of the men here have had experience with the two methods of sampling.

Francis P. Sinn, Palmerton, Pa.—A number of years ago we were using the method of sampling spelter which Mr. Stone spoke of, drilling the slabs. The chief trouble that we had was keeping the drills in shape and in cutting up the strips made by the drills. Our present method of using an ordinary band-saw has saved us a lot of time not only in taking the sample from the slab but the sample requires practically no preparation. It has saved us both time and money. I should say that where we are making 100 tons of spelter a day it saves us the labor of three or four men every day.

George C. Stone.—Mr. Sinn did not mention that he is making the high-grade spelter, of which the analysis has to be within very close limits, and they (New Jersey Zinc Co. of Pa.) take more samples, probably, for the tonnage than at any other works.

E. G. Spilsbury.—What number of samples are supposed to be taken of this high-grade spelter and what percentage of the slabs are taken as samples?

Francis P. Sinn.—In shipping our spelter, we take one slab from every 20 that go in a car, and make three cuts in that slab. We take a cut half through on one side at one end, half through on the same side at the other end, and half through the middle from the other side, so we feel that we cover the slab pretty well in that way, and do not destroy the slab itself by allowing it to break. We are sampling our spelter also as it is made, that is, we sample it hot. We use a small dipper, take a small sample from each ladle, and drop this into water, and the spelter breaks up into a small sort of spatter which is not exactly granulated. In this way we get a sample from every bit of spelter that is made, a little being taken from every ladle that is poured.

Mr. Smith.—I would like to ask Mr. Sinn how he takes the sample from the ladle?

Francis P. Sinn.—We use a small iron dipper and are criticized, of course, by the casual observer. He would say that a certain amount of iron would get into the sample from the dipper. As a matter of fact, a skull forms on the inside of the dipper that protects the sample. We always form that skull in the dipper before keeping the sample. It is really a galvanized dipper.

Mr. Smith.—Is the sample taken from the bottom of the ladle or from the spelter, as poured?

Francis P. Sinn.—Our experience has been that there is no possibility of actually settling the lead in spelter that runs under 1 per cent, of lead, and we are making spelter running considerably under 1 per cent. When we started our method of sampling, we took separate samples from each part of the ladle as it was poured and thoroughly tested out the conditions, but we take the samples from the top of the ladle now. If the spelter runs over 1 per cent, of lead there is a possibility of not getting an accurate sample in this way.

H. L. Glenn, Seattle, Wash, (communication to the Secretary).— I quite agree with the author that in determining the values of bullion the dip samples are better than drills. In order to avoid the possible errors to which he refers, caused by particles of the scorifier-dipper getting into the sample, I suggest the use of a device which we have found comparatively satisfactory in the United States Assay Office in Seattle: Instead of using a scorifier to take the dip samples we use a graphite stirring rod. In the side of this rod near the lower end is bored out a small pocket which will hold an ounce or more of the molten bullion. With this stirring-rod dipper we are able to take the two, dip samples, one from near the top and the other from near the bottom, without delay and with no danger of the difficulties which the author experienced.

Regarding the method of arriving at the top and bottom values of the bar: I prefer to assay each sample separately rather than to mix the samples as suggested by the author. Then, after assaying, if it is desired, an average of the two results may be taken, which secures the same result but still keeps the samples as originally taken, for future use.

The Seattle Assay Office purchases only a small amount of cyanide bullion annually and what we do purchase has, invariably, a larger proportion of silver than is found in the Salt Lake bullion. We have, however, the same trouble as was found by Professor Hance, though the discrepancies in assays are not as great, which is probably accounted for by the larger proportion of silver in the bullion. The Mint Bureau is under obligations to Professor Hance for his painstaking and satisfactory work.

I regret that time will not permit of a more extended discussion of this paper. On the whole the author has handled his subject most admirably and I hope before long that some one may have the time and patience to make further experiments in order to solve some of the problems which Professor Hance was obliged to drop for lack of time.

James H. Hance, Iowa City, Ia. (communication to the Secretary). —Sampling gold-bullion bars according to the same method found so successful with zinc bullion or spelter would doubtless yield much more nearly uniform assay values than drilling methods. For several pertinent reasons I have never tried this method, but I believe it would assist in getting true values, if solid bars only are available. The dip samples, however, are more easily obtained where melting or remelting is done, and, except in rare instances, are sufficiently accurate if carefully prepared and properly used.

In many cases cyanide bars are extremely brittle, and a bar with three slots, each half way through the bar, might be too fragile for ordinary handling. Packing and shipment from assay offices to the mints would also increase the liability of breakage.

Another difficulty, not necessarily serious, is manipulation of such sampling. Since all weights are recorded to hundredths of a troy ounce, all fragments must be saved, and this would be more difficult with saw samples than with drillings. A more serious phase of this would be probable salting or contamination of the sample if the saw were not perfectly clean each time it was used. Cleaning the saw teeth carefully each time would be absolutely necessary, and would take time. The time factor, although not of primary importance, would also be appreciably greater for this method of sampling, and where the bars are numerous and the help limited, this time element might be an important consideration. Finally, I am doubtful if this method would yield results of greater accuracy, if as great, than those of dip sampling.